; Volume Illustration Non-Photorealistic Rendering of Volume Models
Documents
Resources
Learning Center
Upload
Plans & pricing Sign in
Sign Out
Your Federal Quarterly Tax Payments are due April 15th Get Help Now >>

Volume Illustration Non-Photorealistic Rendering of Volume Models

VIEWS: 9 PAGES: 8

  • pg 1
									To appear in Proceedings of IEEE Visualization ’00 (October 2000, Salt Lake City, UT).

            Volume Illustration: Non-Photorealistic Rendering of Volume Models
                                                David Ebert        Penny Rheingans
                                             Computer Science and Electrical Engineering
                                              University of Maryland Baltimore County
                                                        Baltimore MD 21250
                                                  [ebert | rheingan]@cs.umbc.edu



                           Abstract                                   and familiar views of a volume data set, at least for data that has
Accurately and automatically conveying the structure of a volume      an appropriate physical meaning. The second approach is only
model is a problem not fully solved by existing volume rendering      loosely based on the physical behavior of light through a volume,
approaches. Physics-based volume rendering approaches create          using instead an arbitrary transfer function specifying the
images which may match the appearance of translucent materials        appearance of a volume sample based on its value and an
in nature, but may not embody important structural details.           accumulation process that is not necessarily based on any actual
Transfer function approaches allow flexible design of the volume      accumulation mechanism [Levoy90]. This approach allows the
appearance, but generally require substantial hand tuning for each    designer to create a wider range of appearances for the volume in
new data set in order to be effective. We introduce the volume        the visualization, but sacrifices the familiarity and ease of
illustration approach, combining the familiarity of a physics-        interpretation of the more physics-based approach.
based illumination model with the ability to enhance important              We propose a new approach to volume rendering: the
features using non-photorealistic rendering techniques. Since         augmentation of a physics-based rendering process with non-
features to be enhanced are defined on the basis of local volume      photorealistic rendering (NPR) techniques [Winkenbach94,
characteristics rather than volume sample value, the application      Salisbury94] to enhance the expressiveness of the visualization.
of volume illustration techniques requires less manual tuning than    NPR draws inspiration from such fields as art and technical
the design of a good transfer function. Volume illustration           illustration to develop automatic methods to synthesize images
provides a flexible unified framework for enhancing structural        with an illustrated look from geometric surface models. Non-
perception of volume models through the amplification of              photorealistic rendering research has effectively addressed both
features and the addition of illumination effects.                    the illustration of surface shape and the visualization of 2D data,
                                                                      but has virtually ignored the rendering of volume models. We
CR Categories: I.3.7 [Computer Graphics]: Three-Dimensional           describe a set of NPR techniques specifically for the visualization
Graphics and Realism – color, shading, and texture; I.3.8             of volume data, including both the adaptation of existing NPR
[Computer Graphics]: Applications.                                    techniques to volume rendering and the development of new
Keywords: Volume rendering, non-photorealistic rendering,             techniques specifically suited for volume models. We call this
illustration, lighting models, shading, visualization.                approach volume illustration.
                                                                            The volume illustration approach combines the benefits of
                                                                      the two traditional volume rendering approaches in a flexible and
1 Introduction                                                        parameterized manner. It provides the ease of interpretation
                                                                      resulting from familiar physics-based illumination and
For volume models, the key advantage of direct volume rendering       accumulation processes with the flexibility of the transfer
over surface rendering approaches is the potential to show the        function approach. In addition, volume illustration provides
structure of the value distribution throughout the volume, rather     flexibility beyond that of the traditional transfer function,
than just at selected boundary surfaces of variable value (by         including the capabilities of local and global distribution analysis,
isosurface) or coordinate value (by cutting plane).            The    and light and view direction specific effects. Therefore, volume
contribution of each volume sample to the final image is              illustration techniques can be used to create visualizations of
explicitly computed and included. The key challenge of direct         volume data that are more effective at conveying the structure
volume rendering is to convey that value distribution clearly and     within the volume than either of the traditional approaches. As
accurately. In particular, showing each volume sample with full       the name suggests, volume illustration is intended primarily for
opacity and clarity is impossible if volume samples in the rear of    illustration or presentation situations, such as figures in
the volume are not to be completely obscured.                         textbooks, scientific articles, and educational video.
     Traditionally, volume rendering has employed one of two
approaches. The first attempts a physically accurate simulation
of a process such as the illumination and attenuation of light in a   2 Related Work
gaseous volume or the attenuation of X-rays through tissue
                                                                      Traditional volume rendering spans a spectrum from the accurate
[Kajiya84, Drebin88]. This approach produces the most realistic
                                                                      to the ad hoc. Kajiya's original work on volume ray tracing for
                                                                      generating images of clouds [Kajiya84] incorporated a physics-
                                                                      based illumination and atmospheric attenuation model. This
                                                                      work in realistic volume rendering techniques has been extended
                                                                      by numerous researchers [Nishita87, Ebert90, Krueger91,
                                                                      Williams92, Max95, Nishita98]. In contrast, traditional volume
                                                                      rendering has relied on the use of transfer functions to produce
                                                                      artificial views of the data to highlight regions of interest
                                                                      [Drebin88]. These transfer functions, however, require in-depth
                                                                      knowledge of the data and need to be adjusted for each data set.
To appear in Proceedings of IEEE Visualization ’00 (October 2000, Salt Lake City, UT).
The design of effective transfer functions is still an active             •      Volume sample location and value
research area [Fang98, Kindlmann98, Fujishiro99]. While
transfer functions can be effective at bringing out the structure in      •      Local volumetric properties, such as gradient and minimal
the value distribution of a volume, they are limited by their                    change direction
dependence on voxel value as the sole transfer function domain.           •      View direction
     In contrast, there has been extensive research for illustrating
surface shape using non-photorealistic rendering techniques.              •      Light information
Adopting a technique found in painting, Gooch et al. developed a          The view direction and light information allows global
tone-based illumination model that determined hue, as well as             orientation information to be used in enhancing local volumetric
intensity, from the orientation of a surface element to a light           features. Combining this rendering information with user selected
source [Gooch98]. The extraction and rendering of silhouettes             parameters provides a powerful framework for volumetric
and other expressive lines has been addressed by several                  enhancement and modification for artistic effects.
researchers [Saito90, Salisbury94, Gooch99, Interrante95].                      Volumetric illustration differs from surface-based NPR in
Expressive textures have been applied to surfaces to convey               several important ways. In NPR, the surfaces (features) are well
surface shape [Rheingans96, Salisbury97, Interrante97].                   defined, whereas with volumes, feature areas within the volume
     A few researchers have applied NPR techniques to the                 must be determined through analysis of local volumetric
display of data. Laidlaw used concepts from painting to create            properties. The volumetric features vary continuously throughout
visualizations of 2D data, using brushstroke-like elements to             three-dimensional space and are not as well defined as surface
convey information [Laidlaw98] and a painterly process to                 features. Once these volumetric feature volumes are identified,
compose complex visualizations [Kirby99].             Treavett has        user selected parametric properties can be used to enhance and
developed techniques for pen-and-ink illustrations of surfaces            illustrate them.
within volumes [Treavett00]. Interrante applied principles from                 We begin with a volume renderer that implements physics-
technical illustration to convey depth relationships with halos           based illumination of gaseous phenomena. The opacity transfer
around foreground features in flow data [Interrante98]. Saito             function that we are using is the following simple power function:
converted 3D scalar fields into a sampled point representation and
                                                                                   ov = (k os vi ) oe
visualized selected points with a simple primitive, creating an                                         k
NPR look [Saito94]. With the exceptions of the work of Saito
and Interrante, the use of NPR techniques has been confined to            where vi is the volume sample value and kos is the scalar
surface rendering.                                                        controlling maximum opacity. Exponent koe values less than 1
                                                                          soften volume differences and values greater than 1 increase the
3 Approach                                                                contrast within the volume.
                                                                               Figure 1 shows gaseous illumination of an abdominal CT
We have developed a collection of volume illustration techniques          volume of 256×256×128 voxels. In this image, as in others of
that adapt and extend NPR techniques to volume objects. Most              this dataset, the scene is illuminated by a single light above the
traditional volume enhancement has relied on functions of the             volume and slightly toward the viewer. The structure of tissues
volume sample values (e.g., opacity transfer functions), although         and organs is difficult to understand. In Figure 2, a transfer
some techniques have also used the volume gradient (e.g.,                 function has been used to assign voxel colors which mimic those
[Levoy90]). In contrast, our volume illustration techniques are           found in actual tissue. The volume is illuminated as before.
fully incorporated into the volume rendering process, utilizing           Organization of tissues into organs is clear, but the interiors of
viewing information, lighting information, and additional                 structures are still unclear. We chose to base our examples on an
volumetric properties to provide a powerful, easily extensible            atmospheric illumination model, but the same approach can be
framework for volumetric enhancement. Comparing Diagram 1,                easily applied to a base renderer using Phong illumination and
the traditional volume rendering system, and Diagram 2, our               linear accumulation.
volume illustration rendering system, demonstrates the difference              In the following two sections, we describe our current
in our approach to volume enhancement. By incorporating the               collection of volume illustration techniques. These techniques
enhancement of the volume sample's color, illumination, and               can be applied in almost arbitrary amounts and combinations,
opacity into the rendering system, we can implement a wide                becoming a flexible toolkit for the production of expressive
range of enhancement techniques. The properties that can be               images of volume models. The volume illustration techniques we
incorporated into the volume illustration procedures include the
following:
                                                                                 Volume Illustration Rendering Pipeline
      Traditional Volume Rendering Pipeline                                                             volume values f1(xi)
                           Volume values f1(xi)
          shading                                      classification              Volume                   Transfer function
                                                                                   Rendering
     voxel colors cλ(xi)                          voxel opacities α(xi)
                                                                                   Volume Illustration                 Volume Illustration
                shaded, segmented volume [cλ(xi), α(xi)]                           color modification                     opacity modification

                      resampling and compositing
                       (raycasting, splatting, etc.)                                      Final volume sample [cλ(xi), α(xi)]

                            image pixels Cλ(ui)
                                                                                                            image pixels Cλ(ui)

 Diagram 1. Traditional Volume Rendering Pipeline.                            Diagram 2. Volume Illustration Rendering Pipeline.
To appear in Proceedings of IEEE Visualization ’00 (October 2000, Salt Lake City, UT).
propose are of two basic types: feature enhancement, and depth              Figure 3 shows the effect of boundary enhancement in the
and orientation cues.                                                  medical volume. The edges of the lungs and pulmonary
                                                                       vasculature can be seen much more clearly than before, as well as
4 Feature Enhancement                                                  some of the internal structure of the kidney. Parameter values
                                                                       used in Figure 3 are kgc = 0.7, kgs = 10, kge = 2.0.
In a surface model, the essential feature is the surface itself. The
surface is explicitly and discretely defined by a surface model,       4.2 Oriented     Feature    Enhancement:
making “surfaceness” a boolean quality. Many other features,
such as silhouettes or regions of high curvature, are simply               Silhouettes, Fading, and Sketch Lines
interesting parts of the surface. Such features can be identified      Surface orientation is an important visual cue that has been
by analysis of regions of the surface.                                 successfully conveyed by artists for centuries through numerous
     In a volume model, there are no such discretely defined           techniques, including silhouette lines and orientation-determined
features. Volume characteristics and the features that they            saturation effects. Silhouette lines are particularly important in
indicate exist continuously throughout the volume. However, the        the perception of surface shape, and have been utilized in surface
boundaries between regions are still one feature of interest. The      illustration and surface visualization rendering [Salisbury94,
local gradient magnitude at a volume sample can be used to             Interrante95].     Similarly, silhouette volumes increase the
indicate the degree to which the sample is a boundary between          perception of volumetric features.
disparate regions. The direction of the gradient is analogous to             In order to strengthen the cues provided by silhouette
the surface normal. Regions of high gradient are similar to            volumes, we increase the opacity of volume samples where the
surfaces, but now “surfaceness” is a continuous, volumetric            gradient nears perpendicular to the view direction, indicated by a
quality, rather than a boolean quality. We have developed several      dot product between gradient and view direction which nears
volume illustration techniques for the enhancement of volume           zero. The silhouette enhancement is described by:
features based on volume gradient information.

4.1 Boundary Enhancement
                                                                                    (
                                                                          o s = ov k sc + k ss (1 − abs (∇ fn ⋅ V )) se
                                                                                                                        k
                                                                                                                            )
                                                                       where ksc controls the scaling of non-silhouette regions, kss
Levoy [Levoy90] introduced gradient-based shading and opacity          controls the amount of silhouette enhancement, and kse controls
enhancement to volume rendering. In his approach, the opacity          the sharpness of the silhouette curve.
of each voxel was scaled by the voxel's gradient magnitude to               Figure 4 shows the result of both boundary and silhouette
emphasize the boundaries between data (e.g., tissue) of different      enhancement in the medical volume. The fine honeycomb
densities and make areas of constant density transparent (e.g.,        structure of the liver interior is clearly apparent, as well as
organ interiors). We have adapted this idea to allow the user to       additional internal structure of the kidneys. Parameter values
selectively enhance the density of each volume sample by a             used in Figure 4 are kgc = 0.8, kgs = 5.0, kge = 1.0; ksc = 0.9, kss =
function of the gradient. Assume a volume data set containing a        50, kse = 0.25.
precomputed set of sample points. The value at a location Pi is a           Decreasing the opacity of volume features oriented toward
scalar given by:                                                       the viewer emphasizes feature orientation, and in the extreme

              ( )
                                                                       cases, can create sketches of the volume, as illustrated in Figure
     vi = f Pi = f ( xi , y i , z i )                                  5. Figure 5 shows a black and white sketch of the medical dataset
                                                                       by using a white sketch color and making non volumetric
                                                                       silhouettes transparent. To get appropriate shadowing of the
We can also calculate the value gradient ∇f (Pi) at that location.     sketch lines, the shadows are calculated based on the original
In many operations we will want that gradient to be normalized.        volume opacity. Using a black silhouette color can also be
We use ∇fn to indicate the normalized value gradient.                  effective for outlining volume data.
     Before enhancement, voxel values are optionally mapped                 Orientation information can also be used effectively to
through a standard transfer function which yields color value cv       change feature color. For instance, in medical illustration the
and opacity ov for the voxel. If no transfer function is used, these   portions of anatomical structures oriented toward the viewer are
values can be set to constants for the whole volume. The               desaturated and structures oriented away from the view are
inclusion of a transfer function allows artistic enhancements to       darkened and saturated [Clark99]. We simulate these effects by
supplement, rather than replace, existing volume rendering             allowing the volumetric gradient orientation to the viewer to
mechanisms.                                                            modify the color, saturation, value, and transparency of the given
     We can define a boundary-enhanced opacity for a voxel by          volume sample. The use of the HSV color space allows the
combining a fraction of the voxel’s original opacity with an           system to easily utilize the intuitive color modification techniques
enhancement based on the local boundary strength, as indicated         of painters and illustrators. Figure 10 shows oriented changes in
by the voxel gradient magnitude. The gradient-based opacity of         the saturation and value of the medical volume. In this figure, the
the volume sample becomes:                                             color value (V) is decreased as the angle between the gradient and

                 (
     o g = o v k gc + k gs ∇   (    f   )
                                        k ge
                                               )                       the viewer increases, simulating more traditional illustration
                                                                       techniques of oriented fading.
where ov is original opacity and ∇f is the value gradient of the
volume at the sample under consideration. This equation allows         5 Depth and Orientation Cues
the user to select a range of effects from no gradient enhancement     Few of the usual depth cues are present in traditional rendering of
(kgc=1, kgs=0) to full gradient enhancement (kgs >=1) to only          translucent volumes. Obscuration cues are largely missing since
showing areas with large gradients (kgc=0), as in traditional          there are no opaque objects to show a clear depth ordering.
volume rendering. The use of the power function with exponent          Perspective cues from converging lines and texture compression
kge allows the user to adjust the slope of the opacity curve to best   are also lacking, since few volume models contain straight lines
highlight the dataset.                                                 or uniform textures. The dearth of clear depth cues makes
To appear in Proceedings of IEEE Visualization ’00 (October 2000, Salt Lake City, UT).
understanding spatial relationships of features in the volume            relationships in 3D flow data using Line Integral Convolution
difficult. One common approach to this difficulty is the use of          (LIC). She created a second LIC volume with a larger element
hard transfer functions, those with rapidly increasing opacity at        size, using this second volume to impede the view. Special care
particular value ranges of interest. While this may increase depth       was required to keep objects from being obscured by their own
cues by creating the appearance of surfaces within the volume, it        halos. The resulting halos achieved the desired effect, but the
does so by hiding all information in some regions of the volume,         method depended on having flow data suitable for processing
sacrificing a key advantage of volume rendering.                         with LIC.
      Similarly, information about the orientation of features                 We introduce a more general method for creating halo
within the volume is also largely missing. Many volume                   effects during the illumination process using the local spatial
rendering systems use very simple illumination models and often          properties of the volume. Halos are created primarily in planes
do not include the effect of shadows, particularly volume self-          orthogonal to the view vector by making regions just outside
shadowing to improve performance, even though many volume                features darker and more opaque, obscuring background elements
shadowing algorithms have been developed [Ebert90, Kajiya84].            which would otherwise be visible. The strongest halos are
Accurate volumetric shadowing often produces subtle effects              created in empty regions just outside (in the plane perpendicular
which do not provide strong three-dimensional depth cues. As a           to the view direction) of a strong feature.
result, the shape of individual structures within even illuminated             The halo effect at a voxel is computed from the distance
volumes is difficult to perceive.                                        weighted sum of haloing influences in a specified neighborhood.
      We have developed several techniques for the enhancement           In order to restrict halos to less interesting regions, summed
of depth and orientation cues in volume models, inspired by              influences are weighted by the complement of the voxel’s
shading concepts in art and technical illustration.                      gradient. The size of the halo effect is given by:
                                                                                    neighbors h 
5.1 Distance color blending                                                   hi =  ∑
                                                                                    n P −P 2 
                                                                                                n
                                                                                                        (
                                                                                                     1− ∇ (P)
                                                                                                          f i         )
Intensity depth-cuing is a well known technique for enhancing the                            i   n 
perception of depth in a scene [Foley96]. This technique dims            where hn is the maximum potential halo contribution of a
the color of objects far from the viewer, creating an effect similar     neighbor. The haloing influence of a neighbor is inversely related
to viewing the scene through haze. We have adapted this                  to its distance and the tendency of a location to be a halo is
technique for volume rendering, dimming volume sample colors             inversely related to its gradient magnitude.
as they recede from the viewer. In addition, we have augmented                 The maximum potential halo contribution of each neighbor
the standard intensity depth-cuing with a subtle color modulation.       is proportional to the product of the alignment of the neighbor’s
This color modulation increases the amount of blue in the colors         gradient with the direction to the voxel under consideration
of more distant volume samples, simulating techniques used for           (calculated from the dot product between them) and the degree to
centuries by painters, such as aerial perspective [daVinci1506,          which the neighbor’s gradient is aligned perpendicular to the
Beirstadt1881]. This technique exploits the tendency of cool             view direction (also calculated as a dot product). The halo
colors (such as blue) to recede visually while warm colors (such         potential (hn) is given by:
as red) advance.
                                                                                                 (P − Pn )  
                                                                                                                   k hpe
                                                                                 
                                                                                                                           (1 − ∇ (P ) ⋅ V )
     Depth-cued colors start as the voxel color at the front of the
                                                                            hn =  ∇ fn (Pn ) ⋅  i                                       k hse
volume, decreasing in intensity and moving toward the
background color as depth into the volume increases. The                                        P − P                        fn   n

progression of depth-cuing need not be linear; we use an                                          i   n 

exponential function to control the application of depth-cuing.          where khpe controls how directly the neighbor’s gradient must be
The distance color blending process can be described by:                 oriented toward the current location, and khse controls how tightly

               (                    )c
                                                                         halos are kept in the plane orthogonal to the view direction. The
        c d = 1 − k ds d v                   + k ds d v de cb
                             k de                      k                 most strong halo effects will come from neighbors that are
                                         v
                                                                         displaced from the volume sample of interest in a direction
where kds controls the size of the color blending effect, kde            orthogonal to the view direction and that have a large gradient in
controls the rate of application of color blending, dv is the fraction   the direction of this sample.
of distance through the volume, and cb is a defined background                Once the size of the halo effect has been determined,
color. When cb is a shade of grey (cb = (a, a, a) for some value of      parameters control the visual appearance of halo regions. The
a), only standard intensity depth-cuing is performed. Using a            most common adjustment to the halo region is to decrease the
background color that is a shade of blue (cb = (a, b, c) for c > a,      brightness by a scalar times the halo effect and increase the
b), introduces a cool shift in distant regions. Other color              opacity by another scalar times the halo effect. This method
modulation effects are clearly possible, but make less sense             produces effects similar to those of Interrante, but can be applied
perceptually.                                                            to any type of data or model during the illumination process.
     Figure 6 shows the effect of distance color blending. The           Since the halos generated are inherently view dependent, no
ribs behind the lungs fade into the distance and the region around       special processing must be done to keep features from casting a
the kidneys seems to recede slightly. Color blending parameters          halo on themselves.
used in Figure 6 are cb = (0, 0, 0.15), kds = 1.0, kse = 0.5.                 Figure 6 shows the effectiveness of adding halos to the
                                                                         medical volume. Structures in the foreground, such as the liver
5.2 Feature halos                                                        and kidneys, stand out more clearly. Halo parameters used in
                                                                         Figure 6 are khpe = 1.0 and khse = 2.0.
Illustrators sometimes use null halos around foreground features
to reinforce the perception of depth relationships within a scene.
The effect is to leave the areas just outside surfaces empty, even
                                                                         5.3 Tone shading
if an accurate depiction would show a background object in that          Another illustrative technique used by painters is to modify the
place. Interrante [Interrante98] used a similar idea to show depth       tone of an object based on the orientation of that object relative to
To appear in Proceedings of IEEE Visualization ’00 (October 2000, Salt Lake City, UT).
the light. This technique can be used to give surfaces facing the   together with colors from a transfer function. The tone effects are
light a warm cast while surfaces not facing the light get a cool    subtler, but still improve shape perception. The basic tissue
cast, giving effects suggestive of illumination by a warm light     colors are preserved, but the banded structure of the aorta is more
source, such as sunlight. Gooch et al. proposed an illumination     apparent than in a simple illuminated and color-mapped image
model based on this technique [Gooch98], defining a                 (Figure 2). Tone shading parameters used in Figures 7 and 8 are
parameterized model for effects from pure tone shading to pure      kty = 0.3, ktb = 0.3, kta = 1.0, ktd = 0.6.
illuminated object color. The parameters define a warm color by
combining yellow and the scaled fully illuminated object color.
Similarly, a cool color combines blue and the scaled ambient
                                                                    6 Application Examples
                                                                          We have also applied the techniques in the previous sections
illuminated object color. The final surface color is formed by
interpolation between the warm and cool color based on the          to several other scientific data sets. Figures 10 and 11 are volume
                                                                    rendered images from a 256x256x64 MRI dataset of a tomato
signed dot product between the surface normal and light vector.
                                                                    from Lawrence Berkeley National Laboratories. Figure 10 is a
The model assumes a single light source, generally located above
the scene.                                                          normal gas-based volume rendering of the tomato where a few of
                                                                    the internal structures are visible. Figure 11 has our volume
      We implemented an illumination model similar to Gooch
                                                                    illustration gradient and silhouette enhancements applied,
tone shading for use with volume models. As with Gooch tone
shading, the tone contribution is formed by interpolation between   resulting in a much more revealing image showing the internal
                                                                    structures within the tomato. Parameters used in Figure 11 are
the warm and cool colors based on the signed dot product
between the volume sample gradient and the light vector. Unlike
                                                                    kgc= 0.5, kgs= 2.5, kge= 3.0; ksc= 0.4, kss= 500, kse= 0.3.
Gooch tone shading, the illuminated object contribution is                Figure 12 shows a 512x512x128 element flow data set from
                                                                    the time series simulation of unsteady flow emanating from a 2D
calculated using only the positive dot product, becoming zero at
                                                                    rectangular slot jet. The 2D jet source is located at the left of the
orientations orthogonal to the light vector. This more closely
matches familiar diffuse illumination models.                       image and the flow is to the right. Flow researchers notice that
                                                                    both Figures 12 and 13 resemble Schlieren photographs that are
      The color at a voxel is a weighted sum of the illuminated
                                                                    traditionally used to analyze flow. Figure 13 shows the
gaseous color (including any traditional transfer function
calculations) and the total tone and directed shading from all      effectiveness of boundary enhancement, silhouette enhancement,
                                                                    and tone shading on this data set. The overall flow structure,
directed light sources. The new tone illumination model is given
                                                                    vortex shedding, and helical structure are much easier to perceive
by:
                                                                    in Figure 13 than in Figure 12.
                    NL
     c = k ta I G + ∑ (I t + k td I o )
                                                                          Figures 14 and 15 are volume renderings of a 64x64x64
                                                                    high-potential iron protein data set. Figure 14 is a traditional gas-
                     i                                              based rendering of the data. Figure 15 has our tone shading
where kta controls the amount of gaseous illumination (IG)          volume illustration techniques applied, with parameters kty =
included, NL is the number of lights, ktd controls the amount of    0.15, ktb = 0.15, kta = 1.0, ktd = 0.6. The relationship of
directed illumination included, It is the tone contribution to      structure features and the three-dimensional location of the
volume sample color, and Io is the illuminated object color         features is much clearer with the tone-based shading
contribution. Although this model allows for multiple light         enhancements applied.
sources, more than a few is likely to result in confusing images,
since we are not used to interpreting complex illumination
coming from many lights.                                            7 Conclusions
     The tone contribution from a single light source is
interpolated from the warm and cool colors, depending on the        We have introduced the concept of volume illustration,
angle between the light vector and the sample gradient. It is       combining the strengths of direct volume rendering with the
given by:                                                           expressive power of non-photorealistic rendering techniques.
  I t = ((1.0 + ∇ fn ⋅ L) / 2)cw + (1 − (1.0 + ∇ fn ⋅ L) / 2)cc
                                                                    Volume illustration provides a powerful unified framework for
                                                                    producing a wide range of illustration styles using local and
where L is the unit vector in the direction of the light and        global properties of the volume model to control opacity
     cw= (kty, kty, 0),        cc = (0, 0, ktb)                     accumulation and illumination. Volume illustration techniques
describe the warm and cool tone colors. Samples oriented toward     enhance the perception of structure, shape, orientation, and depth
the light become more like the warm color while samples             relationships in a volume model. Comparing standard volume
oriented away from the light become more like the cool color.       rendering (Figures 2, 10, 12, 14) with volume illustration images
     The directed illumination component is related to the angle    (Figures 3, 4, 5, 6, 7, 8, 9, 11, 13, 15) clearly shows the power of
between the voxel gradient and the light direction, for angles up   employing volumetric illustration techniques to enhance 3D depth
to 90 degrees. It is given by:                                      perception and volumetric feature understanding.


                k td I i (∇ fn ⋅ L ) : ∇ fn ⋅ L > 0
                                                                    8 Future Directions
           Io =                                                    We plan on extending our collection of NPR techniques and
                          0 : ∇ fn ⋅ L ≤ 0                         exploring suitability of these volume illustration techniques for
where ktd controls how much directed illumination is added.         data exploration and diagnosis.
     Figure 7 shows modified tone shading applied to the
uncolored medical volume. The small structure of the liver          9 Acknowledgements
shows clearly, as does the larger structures of the kidney. The     We would like to thank researchers        at the Mississippi State
bulges of intestine at the lower right are much more clearly        University NSF Computational Field         Simulation Engineering
rounded 3D shapes than with just boundary and silhouette            Research Center and the Armed Forces      Institute of Pathology for
enhancement (Figure 4). Figure 8 shows tone shading applied         help in evaluating the effectiveness      of these technique and
To appear in Proceedings of IEEE Visualization ’00 (October 2000, Salt Lake City, UT).
guiding our research. We would also like to thank Dr. Elliott           Applications, 18(4), pp. 49-53 (July - August 1998). ISSN
Fishman of Johns Hopkins Medical Institutions for the abdominal         0272-1716.
CT dataset. The iron protein dataset came from the vtk website     [Kajiya84] James T. Kajiya and Brian P. Von Herzen. Ray
(www.kitware.com/vtk.html).       Christopher Morris generated          Tracing      Volume      Densities,   Computer      Graphics
some to the pictures included in this paper. This work supported        (Proceedings of SIGGRAPH 84), 18(3), pp. 165-174 (July
in part by the National Science Foundation under Grants ACIR            1984, Minneapolis, Minnesota). Edited by Hank
9996043 and ACIR 9978032.                                               Christiansen.
                                                                   [Kindlmann98] Gordon Kindlmann and James Durkin. Semi-
References                                                              Automatic Generation of Transfer Functions for Direct
                                                                        Volume Rendering, In Proceedings of 1998 IEEE
[Bierstadt1881] Albert Bierstadt. “Near Salt Lake City, Utah,”          Symposium on Volume Visualization, pp. 79-86.
     Museum of Art, Brigham Young University, 1881.                [Kirby99] R.M. Kirby, H. Marmanis, and D.H. Laidlaw.
[Clark99] John O.E. Clark. A Visual Guide to the Human Body,            Visualizing Multivalued Data from 2D Incompressible
     Barnes and Noble Books, 1999.                                      Flows Using Concepts from Painting, IEEE Visualization
[daVinci1506] Leonardo daVinci, “The Virgin of the Rocks,”              '99, pp. 333-340 (October 1999, San Francisco, California).
     National Gallery, London, 1503-1506.                               IEEE. Edited by David Ebert and Markus Gross and Bernd
[Drebin88] Robert A. Drebin and Loren Carpenter and Pat                 Hamann. ISBN 0-7803-5897-X.
     Hanrahan. Volume Rendering, Computer Graphics                 [Krueger91] Wolfgang Krueger, The Application of transport
     (Proceedings of SIGGRAPH 88), 22(4), pp. 65-74 (August             theory to the visualization of 3D scalar fields, Computers in
     1988, Atlanta, Georgia). Edited by John Dill.                      Physics, pp. 397-406, July 1991.
[Ebert90] David S. Ebert and Richard E. Parent. Rendering and      [Laidlaw98] David H. Laidlaw and Eric T. Ahrens and David
     Animation of Gaseous Phenomena by Combining Fast                   Kremers and Matthew J. Avalos and Russell E. Jacobs and
     Volume and Scanline A-buffer Techniques, Computer                  Carol Readhead. Visualizing Diffusion Tensor Images of the
     Graphics (Proceedings of SIGGRAPH 90), 24 (4), pp. 357-            Mouse Spinal Cord, IEEE Visualization '98, pp. 127-134
     366 (August 1990, Dallas, Texas). Edited by Forest Baskett.        (October 1998). IEEE. Edited by David Ebert and Hans
     ISBN 0-201-50933-4.                                                Hagen and Holly Rushmeier. ISBN 0-8186-9176-X.
[Fang98] Shiaofen Fang and Tom Biddlecome and Mihran               [Levoy90] Marc Levoy. Efficient Ray Tracing of Volume Data,
     Tuceryan. Image-Based Transfer Function Design for Data            ACM Transactions on Graphics, 9 (3), pp. 245-261 (July
     Exploration in Volume Visualization, IEEE Visualization            1990). ISSN 0730-0301.
     '98, pp. 319-326 (October 1998). IEEE. Edited by David        [Max95] Nelson Max. Optical models for direct volume
     Ebert and Hans Hagen and Holly Rushmeier. ISBN 0-8186-             rendering, IEEE Transactions on Visualization and
     9176-X.                                                            Computer Graphics, 1 (2), pp. 99-108 (June 1995). ISSN
[Foley96] James Foley, Andries van Dam, Steven Feiner, and              1077-2626.
     John Hughes, Computer Graphics: Principles and Practice,      [Nishita87] Tomoyuki Nishita and Yasuhiro Miyawaki and
     Second Edition in C, Addison Wesley 1996.                          Eihachiro Nakamae. A Shading Model for Atmospheric
[Fujishiro99] Issei Fujishiro and Taeko Azuma and Yuriko                Scattering Considering Luminous Intensity Distribution of
     Takeshima. Automating Transfer Function Design for                 Light Sources, Computer Graphics (Proceedings of
     Comprehensible Volume Rendering Based on 3D Field                  SIGGRAPH 87), 21 (4), pp. 303-310 (July 1987, Anaheim,
     Topology Analysis, IEEE Visualization '99, pp. 467-470             California). Edited by Maureen C. Stone.
     (October 1999, San Francisco, California). IEEE. Edited by    [Nishita98] Tomoyuki Nishita. Light Scattering Models for the
     David Ebert and Markus Gross and Bernd Hamann. ISBN 0-             Realistic Rendering of Natural Scenes, Eurographics
     7803-5897-X.                                                       Rendering Workshop 1998, pp. 1-10 (June 1998, Vienna,
[Gooch98] Amy Gooch, Bruce Gooch, Peter Shirley, and Elaine             Austria). Eurographics. Edited by George Drettakis and
     Cohen. A Non-photorealistic Lighting Model for Automatic           Nelson Max. ISBN3-211-83213-0.
     Technical Illustration. In Proceedings of SIGGRAPH ’98        [Rheingans96]        Penny Rheingans.          Opacity-modulating
     (Orlando, FL, July 1998), Computer Graphics Proceedings,           Triangular Textures for Irregular Surfaces, Proceedings of
     Annual Conference Series, pp. 447-452, ACM SIGGRAPH,               IEEE Visualization ’96, pp. 219-225 (October 1996, San
     ACM Press, July 1998.                                              Francisco CA). IEEE. Edited by Roni Yagel and Gregory
[Gooch99] Bruce Gooch and Peter-Pike J. Sloan and Amy                   Nielson. ISBN 0-89791-864-9.
     Gooch and Peter Shirley and Rich Riesenfeld. Interactive      [Saito90]       Takafumi Saito and Tokiichiro Takahashi.
     Technical Illustration, 1999 ACM Symposium on Interactive          Comprehensible Rendering of 3-D Shapes, Computer
     3D Graphics, pp. 31-38 (April 1999). ACM SIGGRAPH.                 Graphics (Proceedings of SIGGRAPH 90), 24 (4), pp. 197-
     Edited by Jessica Hodgins and James D. Foley. ISBN 1-              206 (August 1990, Dallas, Texas).
     58113-082-1 .                                                 [Saito94] Takafumi Saito. Real-time Previewing for Volume
[Interrante95] Victoria Interrante, Henry Fuchs, and Stephen            Visualization. In Proceedings of 1994 IEEE Symposium on
     Pizer. Enhancing Transparent Skin Surfaces with Ridge and          Volume Visualization, pp. 99-106.
     Valley Lines, IEEE Visualization ’95, pp. 52-59 (October      [Salisbury94] Michael P. Salisbury and Sean E. Anderson and
     1995, Atlanta GA). IEEE. Edited by Gregory Nielson and             Ronen Barzel and David H. Salesin. Interactive Pen-And-Ink
     Deborah Silver. ISBN 0-8186-7187-4.                                Illustration, Proceedings of SIGGRAPH 94, Computer
[Interrante97] Victoria Interrante and Henry Fuchs and Stephen          Graphics Proceedings, Annual Conference Series, pp. 101-
     M. Pizer. Conveying the 3D Shape of Smoothly Curving               108 (July 1994, Orlando, Florida). ACM Press. Edited by
     Transparent Surfaces via Texture, IEEE Transactions on             Andrew Glassner. ISBN 0-89791-667-0.
     Visualization and Computer Graphics, 3(2), (April - June      [Salisbury97] Michael P. Salisbury and Michael T. Wong and
     1997). ISSN 1077-2626.                                             John F. Hughes and David H. Salesin. Orientable Textures
[Interrante98]     Victoria Interrante and Chester Grosch.              for Image-Based Pen-and-Ink Illustration, Proceedings of
     Visualizing 3D Flow, IEEE Computer Graphics &                      SIGGRAPH 97, Computer Graphics Proceedings, Annual
To appear in Proceedings of IEEE Visualization ’00 (October 2000, Salt Lake City, UT).
     Conference Series, pp. 401-406 (August 1997, Los Angeles,
     California). Addison Wesley. Edited by Turner Whitted.
     ISBN 0-89791-896-7.
[Treavett00] S.M.F. Treavett and M. Chen. Pen-and-Ink
     Rendering in Volume Visualisation, Proceedings of IEEE
     Visualization 20000, October 2000, ACM SIGGRAPH
     Press.
[Williams92] Peter L. Williams and Nelson Max. A Volume
     Density Optical Model, 1992 Workshop on Volume
     Visualization, pp. 61-68 (1992). ACM.
[Winkenbach94] Georges Winkenbach and David H. Salesin.
     Computer-Generated Pen-And-Ink Illustration, Proceedings
     of SIGGRAPH 94, Computer Graphics Proceedings, Annual
     Conference Series, pp. 91-100 (July 1994, Orlando, Florida).
     ACM Press. Edited by Andrew Glassner. ISBN 0-89791-
     667-0.




                                                                                         Figure 12. Atmospheric volume rendering
                                                                                         of square jet. No illustration enhancements.




Figure 1. Gaseous illumination of medical      Figure 5. Volumetric sketch lines on CT
CT volume. Voxels are a constant color.        volume. Lines are all white.
To appear in Proceedings of IEEE Visualization ’00 (October 2000, Salt Lake City, UT).




Figure 2. Gaseous illumination of color- Figure 3.         Color-mapped gaseous     Figure 4. Silhouette and     boundary
mapped CT volume.                        illumination with boundary enhancement.    enhancement of CT volume.




 Figure 6. Distance color blending and Figure 7. Tone shading in CT volume. Figure 8. Tone shading in colored volume.
 halos around features of CT volume.   Surfaces toward light receive warm color. Surfaces toward light receive warm color.




                                             Figure 10.      Standard atmospheric Figure 11. Boundary and silhouette
                                             volume rendering of tomato.          enhanced tomato.

 Figure 9. Orientation fading.    Surfaces
 toward viewer are desaturated.




Figure 13. Square jet with boundary and Figure 14. Atmospheric rendering of        Figure 15. Tone shaded iron protein.
silhouette enhancement, and tone shading. iron protein.

								
To top